New Interface for ”Wetten.nl”
Bart H.L. Vredebregt6229778
Bachelor thesis Credits: 18 EC
Bachelor Opleiding Kunstmatige Intelligentie University of Amsterdam Faculty of Science Science Park 904 1098 XH Amsterdam Supervisors dhr. dr. R.G.F. Winkels dr. A.W.F. Boer Leibniz Center for Law University of Amsterdam
Faculty of Law Vendelstraat 8 1000 BA Amsterdam
Summary
In this project a new improved interface for a legislation portal like “wetten.nl” is created. As this portal lacks a proper suggestion system and does not con-tain any case law. The interface created in this project features a suggestion system which suggests relevant articles for the currently selected article based on references in both legislation and case law. As well as the ability to view case law documents within the same interfae. We have focused on users that have experience with legal system, as an interface for inexperienced users would require further explanation of the relations.
The references from legislation were retrieved from the Metalex document server. The references from case law were extracted from documents using regular expressions inside a Python custom pre-processing tool allowing us to simultaneously extract references as well as other data about case law document allowing for further research into other aspects of case law. A network of refer-ences was created in order to find related legislation and case law for a single node through betweenness centrality based on a local network.
Contents
1 Introduction 5
2 Earlier work 6
3 Domain selection 7
4 Reference extraction from case law 9
4.1 Locating references . . . 9 4.2 Resolving references . . . 10 4.3 Evaluation . . . 11 4.3.1 Locating evaluation . . . 11 4.3.2 Resolving evaluation . . . 12 5 Interface 13 5.1 Suggestion system . . . 13 5.1.1 Network creation . . . 14 5.1.2 Network analysis . . . 16 5.2 Historical documents . . . 17 5.3 Interface Evaluation . . . 19 6 Conclusion 20 7 Discussion 21 8 Future work 22 8.1 Reference locating . . . 22 8.2 Reference resolving . . . 22
8.3 Alternative uses of the data . . . 23
8.4 Interface . . . 23
9 References 25
A SPARQL query used to retrieve reference for legislation 26
Acknowledgements
I would like to thank Radboud Winkels and Alexander Boer, as they both provided me with many great ideas and gave me advice whenever it was required. I also would like to thank my fellow student, Alexander van Someren, for his part in creating the reference parser for case law and for his support during the project.
1
Introduction
Any search for information about Dutch legislation will most likely arrive at the official portal of the Dutch government namely ”Wetten.nl”. One weakness of this portal is the absence of a suggestion system, making it almost impossible to find all related legislation regarding any particular subject. Previous attempts to improve the portal included trying to add a feature which suggests alternative articles based on references contained within the laws. However, we propose an suggestion system based on references found in laws in combination with references found in case law referring to law articles, thus combining the best of both practice and theory. Leading to the question “Is it possible to improve the relevance measure for law articles by adding case law?”.
In this project we have focused on users that have experience with legal system especially with reading law as this ensures the user knows the importance of references. Another advantage is the fact that an experienced user is able to find a relevant starting point for its search for information. An interface for inexperienced users would require the interface to assist in finding a suitable starting pointing, changing its tasks from given suggestions based on articles to the task of matching real life situation to an article.
In order to be able to use reference from case law we need to enhance an existing database containing transcripts of case law that are deemed to be rel-evant1 for future reference. The enhancement focuses on two aspects: finding references within the text and resolving the located references to the correct article within the legislation. This enhancement is necessary due to the fact that Dutch case law has no annotated references. Thus we are trying to answer the question: “Is it possible to automatically find and resolve references in case law?”.
From this enhanced database we will extract references which can be placed in a network together with the references within laws. This network is the key to our suggestion system since it allows us to calculate a relevance measure for articles through network analysis techniques which should improve the existing relevance measures used in earlier work.
1Case law in the Netherlands is only published when it is either selected by a commercial
2
Earlier work
In Winkels et al. (2013)[1] a suggestion system was build based on references between articles within Dutch legislation using the Metalex Document server as described in Hoekstra (2011) [2]. Similar to Winkels et al. (2013)[1] Mazzega et al. [3] applied network analysis to French legislation. In Fowler (2003) [4] a network was created from court orders of the U.S. Supreme Court in order to identify most important Court precedents. We propose to combine both case law and legislation in order to generate more relevant suggestions.
In Opijnen (2014) [5] is suggested that references in case law should be presented in a standardised fashion among other suggestions that would improve the accessibility of case law in the Netherlands, both for humans and computers. As dutch case law lacks a proper standard for references to legislation and case law.
3
Domain selection
In earlier work from Winkels et al. (2013)[1], on which this project is based, the domain was limited to tax law in order to keep the amount of legislation in control. In this project we faced this choice once again. If the tax law domain was used we could reuse certain aspects of the earlier project but it would limit us to a relative small and enclosed domain. Another option was using the foreigners law domain as it is larger and has had a lot of publicity in the recent years. Another advantage of the foreigners law domain is the fact that the Leibniz Center for Law were working on projects for the IND2recently
making finding users for the evaluation easier.
The most important aspect of a domain for this project is the amount of published case law within the particular domain therefor an analyse was made to find out which domain was most suitable for this project. In figures 1 and 2 the distribution of publication and total number of cases for each domain is shown for the period 1990-2013. As mentioned before, case law in the Netherlands is only published when it is either selected by a commercial publisher or a government agency. Cases are selected based on the relevance of the case for future cases.
Within in the tax law domain their is an slight increase in publication within the recent year as is visible in figure 2. Within the foreigners law domain the amount of publications is constant around 1.000 within the period 2007-2013. Another benefit of the foreigners law domain is the fact that only a small percentage is published meaning that only the most relevant cases are published whereas the tax law domain has a publication rate of almost 25% which means we would have to filter out irrelevant cases before hand in order to prevent irrelevant cases from disrupting the results. As the tax domain has may cases that are relevant on specific details in the situation of the people involved. Thus making the foreigners law domain the domain of choice.
Figure 1: Distribution of case law publication among years within the foreigners law domain.
Figure 2: Distribution of case law publication among years within the tax law domain
4
Reference extraction from case law
In order to enable us to use references found in case law for the suggestion system we need to acquire these references. In the publicly available official database (rechtspraak.nl) references are only recently added on a small scale for recent cases. These references are most of the time a small fragment of the references contained in the text and are not linked to a location in the case laws text. As can be seen at line 17 of appendix B. These properties make them unsuitable for our suggestion system as we need as many as possible references and preferably also the location within the text in order to help the user recognise the references, thus we need to acquire this data through our own pre-processing application. In this chapter we will describe how we locate the references followed by a section about how we resolved them. Concluding with a section describing the evaluation process.
4.1
Locating references
Firstly we need to locate references this is done through a regular expression which return not only the reference string but also extracts relevant parts of the references e.g. article number and law title. Regular expressions were chosen based on their predictable behaviour and because of their ability to find struc-tured variable data aspects that are both favourable for the current project. Alternatively learning algorithms, like SVMs, could have been used but these require large annotated data which were unavailable for this project, thus mak-ing regular expressions the solution of choice.
In order to locate reference it is required the reference starts with certain keywords which indicate the start of a reference e.g. “Artikel”3and “Boek”4.
These keywords in combination with the requirement that the keywords are trailed by a number make a good indicator for a references.
Using this knowledge the complete regular expression in figure 3 was cre-ated. In the expression line 2 till 6 are used to capture the article (or book or chapter) number and the connected number which in many cases contains also the paragraph number. Line 8 till 15 are used for capturing a more specific position within the article which are not used in this project but excluding them from the expression would cause loss of the rest of the reference. Finally lines 17 till 26 capture the law title and any information in parentheses trailing it as it frequently holds additional information about the reference.
This regular expression matches most of the references but does not match reference towards the “Vreemdelingencirculaire” as it uses a completely differ-ent format compared to other references. Normal references are formatted like “Article X of Law Y” whereas the “Vreemdelingencirculaire” uses a format like “Law Y chapter X/Z”. To illustrate the problem an example of a reference to-wards the “Vreemdelingencirculaire” “Vreemdelingencirculaire 2000 (Vc 2000) hoofdstuk B10/2.4”.
3Article 4Book
1 ([^a-zA-Z] 2 (?: 3 (?:[Aa]rtikel|[Aa]rt\\.?) ([0-9][\(0-9a-zA-Z:.\)]*) 4 |[Bb]oek ([0-9][\(0-9a-zA-Z:.\)]*) 5 |[Hh]oofdstuk ([0-9][\(0-9a-zA-Z:.\)]*) 6 ),? 7 ( 8 (?:\s+
9 (?:lid|aanhef en lid|aanhef en onder|onder)?
10 (?:[0-9a-z ]|tot en met)+,?
11 |,? (?:[a-z]+ lid|[a-z]+ en [a-z]+ lid),?
12 )*
13 )
14 (,? onderdeel [a-z],?)? 15 (,? sub [0-9],?)?
16 (?:
17 (?: van (?:de|het|)(?: wet)?|,?)?]\s*
18 ((?:
19 (?:wet|bestuursrecht|Wetboek van 20 |op het [A-Z0-9][a-zA-Z0-9]* 21 |[A-Z0-9][a-zA-Z0-9]*) 22 (?:[^\S\n]*|\\.) 23 )+) 24 )?\s* 25 (?:\(([^\)]+?)\))? 26 )
Figure 3: Regular expression for locating references inside case law.
4.2
Resolving references
After the references are located we need to resolve them to a reference to the correct law. The resolving is done through comparing the found title to a large database which holds for each law the official title but also citation title(s), unofficial title(s) and abbreviations. Both the found title and the titles in the database are normalised, stripped of unnecessary white-spaces and converted to lowercase, in order to find all possible solutions.
The resolving processes sometimes results in multiple possible solutions most of which are caused by shared abbreviation. An example of a ambiguous law title is ‘WAV‘ which is a abbreviation of ”Wet Arbeid Vreemdelingen”5but also
of ”Wet Ammoniak en Veehouderij”6 and ”Wet Ambulancevervoer”7. In order
to solve this problem we use earlier references from the same text. This is done by checking whether a ambiguous reference was found before in the text as most of the time before a abbreviation is used in a reference the full name is used
5Labor Immigration Law 6Ammonia and Livestock law 7Ambulance Act
in an earlier reference. Thus we can use the knowledge of the earlier use to indicate how the reference should be resolved.
During the locating of the references we also find references without a title e.g. ”Artikel 1 stelt”8. These references are resolved through the assumption
that if there is no title in a reference it is using the same law as the previous ref-erence. Using this approach we are also able to resolve references that explicitly refer to the previous reference e.g. ”Artikel 5 van deze wet”9. Addiontailly we
added a blacklist which allows us to specifically reject found law titles that are clearly not a law title, e.g. ”Nederland” or ”Onze minister”10, and treat them as if there was no title found. Thus giving us the ability to slightly correct small mistakes made during the location processes.
The resulting references are added to the XML of the case law document as can be seen at line 30 of appendix B. References added by our parser can be recongized by “metaLexResourceIdentifier” attribute of the
“dcterms:reference”11tag and the fact is has a “dcterms:string” as child.
4.3
Evaluation
In order to evaluate reference extraction process we used a two step evaluation. First we evaluated the locating process and finally we evaluated the resolving process.
4.3.1 Locating evaluation
True Positive: 141 False Positive 1 False Negative: 22 True Negative: 0
Precision: 0, 9930 Recall: 0, 8712 Table 1: Confussion matrix for locating references
The evaluation of the locating process was done by randomly selecting 25 case law document within in these document references were marked by hand. By comparing these results to the results of our locating process we are able to calculate both recall and precision as shown in table 1.
In this evaluation true positive is a reference which was found in both pro-cesses 12 and false positive a reference which is found in the locating process but not in the evaluation. False negatives are references that are found in the evaluation but not found in the locating process.
As is clearly visible in the results of the evaluation the precision is high, almost 1, which is perfect for our task as it is better to miss some references
8Article 1 states 9Article 5 of this law 10Our minister
11Due to the namespaces in XML dcterms is replaced by ns1 in the appendix 12Locating and the evaluation
then to add non existence references. This is a design choice as for some project it is better to have more false positive then false negatives.
4.3.2 Resolving evaluation
In order to evaluate the resolving of located references a sample of 250 references was randomly selected from all the located references.
For each references there were 4 options; true positive if the resolved article matched the article it is supposed to be; false positive if a reference was resolve which should not be resolved e.g. European Union treaties as they are outside the scope of this project. False negative are references which are not resolved but should have been resolved and True negative are references that are not resolved and should not have been resolved.
Again the results show a high precision as it is better to not resolve a refer-ence then to resolve it incorrectly. As most users prefer a smaller amount but correct resolved references over a larger amount but resolved incorrectly.
True Positive: 198 False Positive 10 False Negative: 36 True Negative: 6
Precision: 0, 9519 Recall: 0, 8462 Table 2: Confusion matrix for resolving references
5
Interface
In the previous interface, as described in Winkels et al.[1], users stated the interface contained to much information. Thus during the design of this interface simplicity and clarity were the main focus. Resulting in a modern looking interface, as can be seen in figure 4, in comparison to the more cluttered previous interface, as can be seen in figure 5.
Figure 4: Image showing the actual interface of this project.
Figure 5: Image showing the interface of earlier work.
5.1
Suggestion system
The main focus of this project is of course the improvement of the suggestion system through the use of references in both legislation and case law. In the
interface the suggestions are located in at the left side of the screen as indicated in figure 6. In this chapter we will explain how the network, on which the suggestions are based, in build and how we analyzed the results in order to return correct suggestions.
Figure 6: Image showing the suggestions generated by the system in red.
5.1.1 Network creation
The network used for our suggestion system consist of information coming from two sources namely; the Metalex document server and our own reference extrac-tion script. As each has different features we will describe them in two separate paragraphs. Each describing the way a local network in created based on the input of an expression level article13.
Case law The case law network is created using the references extracted in our extraction script. As these references are not unique it allows us to count them and thus create weighted edges from a ECLI14 towards articles or either
laws. The weights of the edges are calculated using the following formula:
W = 1 n
Where n is the amount of occurrences of a certain reference (from a case law document toward a article) and W is the weight of a edge. The lower the weight the stronger the impact on the network of the edge is.
13The expression level of a law or an article is identifier which includes the version of the
document, whereas the work level just contains the general identifier (BWB)
All the references in this network uses the work level of legislation, this means it holds no version information. This is done because we have no way of knowing towards which version of a law the judge was referring. Meaning the input, which is at the expression level, is converted to work level to be able to match it with nodes within the network.
The network that was used in this project containing only references of case within the Foreigners law domain consist of 85.639 edges. The size of the file containing the edges has a size of around 9.7 MB and is formatted in a space separated format for example an edge with weight 1 will be represented as “<node1> <node2> 1.0”.
Legislation The legislation network was created using the Metalex Document Server[2]. As Metalex contains information about all the Dutch legislation a SPARQL query was written to fetch only the references containing a BWB found in the case law network. The full SPARQL query can be found in appendix A. The query was run at the endpoint of the Metalex Document Server15 and
exported as “TSV”.
The resulting document contains two columns, one containing the URL for the reference object and one containing the target of the reference. The target of the reference is always at the work level but the reference object URL is neither work level or expression level at this point. But the reference object URL does contain the expression level of the source as can been seen below:
http://doc.metalex.eu/id/BWBR0012002/hoofdstuk/2/artikel/2.1/ lid/1/al/2/extref/1/nl/2011-11-01
The source of the example above is:
http://doc.metalex.eu/id/BWBR0012002/artikel/2.1/2011-11-01
This source can easily be extracted using the following two regular expres-sions. This can be done in a single regular expression but for the sake of clarity we have chosen to do it in two expressions.
First regular expression:
^"http://doc.metalex.eu/id/(BWB[VR]\d{7}) [^"]*?/(\d{4}-\d{2}-\d{2})".*?"(.*?)"$
Replace \n in this string with the n-th matching group: <http://doc.metalex.eu/id/\1/\2> <\3>
Second regular expression:
^"http://doc.metalex.eu/id/(BWB[VR]\d{7})[^"]*?
/artikel/([^/]+)[^"]*?/(\d{4}-\d{2}-\d{2})".*?"(.*?)"$ Replace \n in this string with the n-th matching group:
<http://doc.metalex.eu/id/\1/artikel/\2/\3> <\4>
Sometimes spaces sneak into the URLs which means we have to encode them in order to prevent them from interfering with our format. This is done by replacing anything that matches the following expressions “[^>]\s[^<]” with “\%20”16. This will replace any space that is not used to separate the URL of
the source and the target.
The resulting file contains 100.000 edges and is sized 10.8 MB. The fact that it contains exactly 100.000 edges is most likely a consequent of a security feature of Metalex’s SPARQL endpoint.
5.1.2 Network analysis
When a user clicks an article in order to get suggestions several steps are taken. In this section we will explain the process and the design choices made for each step.
Parsing input When a user click an article an AJAX request is sent to the server containing the law that is currently viewed17 and the selected article. First step is to combine all the information and create both the work level and the expression level identifier for the current node in order to be able to locate it in the networks.
Selecting local network (Case law) The first step is to check whether the current node appears in the case law network, if not we should continue to the next step. If it appears we can start creating a network. In order to keep the computation power required for finding suggestions with a reasonable range we use only a small piece of the network for the analysis. This is done through the creation of a so called ego graph a local network containing all the nodes and edges within a certain distance around the current node.
As the case law network only consist of case law to legislation reference we know that if we take 2 steps (ignoring the direction of the reference) we will end up in a node of the same type as the one we started in. Thus a distance of 2 givens us all the laws related to the current node. But this can result in huge networks which are still problematic in terms of computability. Therefor start searching at a distance of 0.4 giving us a smaller network consisting of only the closest nodes. Because we want to suggest legislation we then have to check whether there are more the N law nodes in the network18 before we continue.
If this is not the case we try larger distances19 until we have enough law nodes
in our network. If even at a distances of 2.0 there are not enough law nodes we use the largest network found.
16a single space in a URL is always encoded this way 17This is always at the an expression level
18Excluding the current node
Selecting local network (Legislation) Again the first step in to check whether the current node is in the legislation network. When it is we again create a local network using ego graph. But as the legislation network is not weighted we cannot reduce the size of the network by increasing the distance from the current node. Yet there is a factor that slightly reduces the size of the local network, the legislation network contains expression level nodes thus it allows us to only select references coming from the selected version of the current node.
Downside to the fact that the legislation network contain expression level nodes is that they don’t match with the nodes found in the case law network. Therefor we have to rename all the expression level nodes in the local legislation network to their work level equivalents.
Combining networks In the best case we now have both a local case law network and a local legislation network thus we need to combine them. Use the NetworkX library in Python we can add all the nodes from the one to the other creating one large network but as the case law is weighted and the legislation is not we have set the weight of edges from the legislation network to a certain value. We chose for a value of 0.1 as it allows the legislation to influence the result but not over rule the case law references. If we have either a local case law network or a local legislation network we can just use that network. If no network could be found we can safely assume that we cannot give suggestions for this node.
Selecting suggestions In order to select the most relevant articles from the local network we need to use some method to rate the relevance of an article in a network. As we have a local network which is formed around our current node we hypothesised that calculating the betweenness centrality for each node using NetworkX betweenness_centrality function and ordering them would result in a ranking with the current node as highest scoring node. This hypothesis proven to be false as the current node was not always the number 1 yet always appeared in the top 5. Thus the ranking does give useful information about the importance of nodes in our network. From the ranking we then create a list of best suggestions for both legislation and case law.
A larger important optimisation was the use of the betweenness centrality approximation as it allows us to estimate instead of calculated the betweennes centrality result in a drop in calculation time of up to 50% without significant changes in the result.
5.2
Historical documents
In order to further improve the usability of the interface we allow users to switch to other versions of laws. Currently this is done rather naive by letting the user select a version and return him the law text of that version. The improvement which was planned for this project was to indicate to the user how much articles have changes over the course of the versions. But this part couldn’t be realised due to unforeseen problems with Metalex.
Figure 7: Image showing the version history selection in red.
In order to be able to indicate how much an article has changed among versions we need to retrieve all the versions of the text. With these versions of the article we can compare two articles with each other through techniques like for example Dynamic Time Warping in order to see what changed between the version and what did not.
The problem with the current version of Metalex is the fact that the refer-ences to article use long URLs containing a lot of information making it impos-sible to easily track down other version of an article if some of the information in the URL changes. For example ”Artikel 3.4 van het Vreemdelingenbesluit”20
has the following URL in the version of 1-1-2012:
http://doc.metalex.eu/id/BWBR0011825/hoofdstuk/3/afdeling/2/ paragraaf/1/artikel/3.4/nl/2012-01-01
But at some point in time the structure of the document has changed, an sub paragraph was added, causing the URL for the article in the version from 1-2-2014 to change to this:
http://doc.metalex.eu/id/BWBR0011825/hoofdstuk/3/afdeling/2/ paragraaf/1/sub-paragraaf/1/artikel/3.4/nl/2014-02-01
As can be seen in the above example the change of the structure of the document can easily break the link between 2 versions of an article. This can be solved by changing the way Metalex stores it articles to a combination of BWB, article number and version date. Alternatively a OWL:SameAs relation could be defined between the current article URLs and a new URL containing just the BWB, article number and version date.
When this problem is solved proper recognition of changes in article can be realised. Currently the JavaScript in our interface already uses a combination of the laws expression level and the article number in order to return the user to the correct article independent of changes in paragraphs. Which could be seen as a prove of concept for the suggested identifier for articles.
5.3
Interface Evaluation
In order to evaluate the interface we used a questionnaire with both formative and summative questions. In total three persons responded to our request to evaluate the interface.
The users appreciate the clear and uncluttered layout of the interface, they indicate that it is easily understandable without the need for a help page. This is a positive result as one of the focuses of the interface was to create a clear interface.
Another positive point mentioned by a user is the indication of the amount of relevance references towards legislation in the case law overview. This feature might needs to be added in the legislation overview.
Weak points are the loading times, these are too long. Which is a partial result of the current hosting server but also of the choice to do calculations in real time.
About the network overview function one user mentioned is looks nice but is impractical due to the need to click on nodes to see what is represents. Another user mentioned he would like to have a search function. But this could be considered as being outside the scope of the project.
Users respond positive to the suggestion based on case law, they indicate it adds relevant articles that where missing in the previous interface. But they also indicate that the suggested relevant case law sometimes is not really a part of the legal foundation of the case.
One user noticed the fact that we are not able to located references to-wards the “Vreemdelingencirculaire” as mentioned in 4.1. Another states that it appears to miss some references in the legislation but it seems to be almost completed in case law.
6
Conclusion
In this project we have shown that it is indeed possible to automatically locate and resolve references in case law. The resulting parser is not perfect as it misses some types of references but has an high enough precision to be useful in this project and for further research. Thus we could concluded that it is possible to automatically find and resolve references in case law.
With the necessary adjustment the results of the parser could be improved further. We now have created a foundation for further research into references in case law on a larger scale due to the automatic extraction which allows to extract information from large corpora.
The created network of references currently uses betweenness centrality to find related articles and case law based on a single input node. Resulting in improved suggestions compared to the results of earlier work, thus we can con-cluded that it is possible to improve the relevance measure for law articles by adding case law. The software used to locate the references was purposely build in a way it allows for easy experimenting with other network analysis techniques, thus allowing for further research.
The interface created has a clear uncluttered lay out which is appreciated by users, the references in both legislation and case law need some improvements. The network overview function is nice in order to get a global idea of the network but is impractical at this point especially for large networks.
7
Discussion
During this project several choices were made that affected the outcome of this project. Firstly we decided to use regular expressions which are show predictable behaviour and are perfect for finding fixed patterns in data. But this is also it weakness, we now only find those references that follow the predictable pattern we have defined beforehand. Thus we will never discover new forms of references making the results of the pre- processing just a simple assistant to help a reader to do what it already could do. Switching to a learning algorithm for instance would help us find other references that we could not easily detect with our current system. But the results have proven that we are currently finding such a large percentage of the references we can safely conclude that regular expressions are the best suited approach at this point.
Another major design choice was to do the suggestion calculation in real time instead of using pre-processing. Most similar projects, like the interface create in Winkels et al. (2013)[1], tend the prefer pre-processing above real time calculations as it gives the user the experience that everything is done instantly. We thought about doing the same but we realised that doing so is not realistic when using the full legal system if we would have to compute and store every possible node beforehand we would need an insane amount of storage. Searching through these amounts of data in order to find the results may even take longer than to compute them real time based on domain specific networks which.
For example the current network consists of 2 file of each around 10 MB totalling at 20MB21 for the complete domain of Foreigners Law. The current
web application requires less then a second to come back with result when it is not in the network but takes around 7 seconds to return suggestions when handling large local networks. But note that this is done on a single core virtual private server which are notorious for there low performance. This shows it is realistic to have servers pre-load the networks of one or more domains allowing it to do calculation in almost real time within several years.
As seen in section 5.3 the interface was evaluated by in total three persons which is not enough to base claims on. This was caused by technical difficulties during the deployment of the interface to the world wide web causing a delay of several days. But we assume that these three users represent a significant part of the user base in order to be able to say thing about the interface.
8
Future work
During each stage of the process several possibilities for further improvements and research surfaced. Below these possibilities will be describe as detailed as possible.
8.1
Reference locating
The reference locating procedure could be improved in several ways; Firstly the regular expression could be improved to enable for more precise matches of the complete reference or a smaller regular expression could be written that focuses on finding the starting point of a references followed by another technique to determine where the references ends.
As references have a reasonably fixed pattern in them SVM could be used to locate them. The problem with this approach is of course the need for a large annotated corpus which could be solved by using the corpus retrieved in this project after human correction. Another advantage of our corpus is that it contains both the references string and the paragraph the string is in allowing for small scale testing and easier human readability.
Also the problem mentioned in section 4.1 with the “Vreemdelingencircu-laire” has to be solved. As this accounts for a small part but important part of the missed references within the Foreigners law domain.
8.2
Reference resolving
Especially in the reference resolving stage many thing could be improved. Firstly the ambiguity problem could be tackle in several alternative ways to improve the results. The biggest problem is the problem with ambiguity of abbreviations as they cause a large part of the incorrect resolved references. This problem could be solved by creating a domain specific law title database as the abbreviations tend not to overlap within a domain, only with abbreviations outside its domain. Luckily the meta data from the case law has a field containing the domain of the case making this a viable option.
Another alternative way to resolve some of the ambiguity is through validat-ing the location a reference is pointvalidat-ing at. For example, we have a references that points to article 16a of a law, but we don’t know which law as the resolving process resulted in two options. In that case we could validate whether both laws have a article 16a, if one has this article and the other has not the am-biguity is solved. This is a simple example but some reference contain a more positioning information like e.g. ”artikel 126aa, tweede lid”.
A final alternative to resolve some ambiguity is by using announcements of a new abbreviation e.g. “artikel 83 van de Vreemdelingenwet 2000 (hierna: de Vw 2000)”22. By storing these announced abbreviation it becomes easier to resolve
any found ambiguous abbreviation as the abbreviation is now defined local.
Another problem that occurs during resolving are references without a law name e.g. ”artikel 1 stelt”. These are currently resolved by assuming that it is referencing the previous referenced law. But this isn’t always true, it could be article 1 always refers to law A and article 6 to law B causing our implementation to assume it is referencing to the previous law but in fact referencing to another law. So a possible improvement could be achieved by storing if a article is used before with a certain law enabling the resolving process to resolve this kind of references.
An addition to the current resolving process would by adding version in-formation to the references using the meta data provided in the header of an case law document23. For example the “Uitspraakdatum”24 could be used to decided which version mostly likely was referred to at during the trial.
8.3
Alternative uses of the data
During the extraction we also extracted other data which has no use in this project. An example of this data is a ordered list of reference occurrences in a case law document. This data could for example be used to find a structure in these documents where certain articles are always referred to first and other always appear in the latter part of the document.
Another use of this data is creating a list a N-grams to help predict the next references. For bi-grams this would mean that if we have a set of bi-grams and a previous reference we could look up in our bi-gram set which reference is most likely to come next enabling us to create a auto complete system for references.
8.4
Interface
As mentioned before in section 5.2 the history or time travel option of the interface needs to be improved with a feature which allows users to see the amount of change in a article between versions.
Another improvement to the interface would be to store a history of visited relevant articles and use this in order to improve the suggestions as it allows us to use more than one node as a starting point of the network. Thus creating a network that has nodes that are potentially even better suggestions as we now can look for the nodes connecting the historical nodes, as shown in figure 8 instead of searching for a related node. Just like the recommendation systems all around us which uses the history of the user to be able to give relevant suggestions. Making the suggestion system less node specific but more user session specific.
23An example of an case law document can be found in Appendix B. 24Ruling date.
Historical article 1 Historical article 2
Ideal suggestion
Current article
References References
References
Figure 8: Simple network showing the position of the ideal suggestion when using article history
9
References
[1] R. Winkels, A. Boer, and I. Plantevin, “Creating context networks in dutch legislation,” Frontiers in Artificial Intelligece and Applications, vol. 259, pp. 155–164, 2013.
[2] R. Hoekstra, “The metalex document server,” in The Semantic Web – ISWC 2011 (L. Aroyo, C. Welty, H. Alani, J. Taylor, A. Bernstein, L. Kagal, N. Noy, and E. Blomqvist, eds.), vol. 7032 of Lecture Notes in Computer Science, pp. 128–143, Springer Berlin Heidelberg, 2011.
[3] P. Mazzega, D. Bourcier, and R. Boulet, “The network of french legal codes,” in Proceedings of the 12th international conference on artificial intelligence and law, pp. 236–237, ACM, 2009.
[4] J. H. Fowler, T. R. Johnson, I. Spriggs, F. James, S. Jeon, and P. J. Wahlbeck, “Network analysis and the law: Measuring the legal importance of supreme court precedents,” Political Analysis, vol. 15, no. 3, pp. 324–346, 2007.
[5] M. van Opijnen, OP EN IN HET WEB - Hoe de toegankelijkheid van rechter-lijke uitspraken kan worden verbeterd. Boom juridische uitgevers, 2014.
A
SPARQL query used to retrieve reference for
legislation
1 PREFIX mo: <http://www.metalex.eu/schema/1.0#> 2
3 SELECT DISTINCT ?ref, ?bwbs WHERE 4 { 5 { 6 ?ref mo:cites ?bwbs. 7 FILTER (regex(str(?bwbs),"(?:BWBR0011825|BWBR0009876| BWBR0005537|BWBV0001000|BWBR0015703|BWBR0011823| BWBR0007333|BWBV0002508|BWBR0017646|BWBV0004701| BWBR0002638|BWBR0001903|BWBR0005181|BWBV0001002| BWBR0006723|BWBR0001854|BWBR0012002|BWBR0017959| BWBR0005310|BWBR0005290|BWBR0007149|BWBR0011468| BWBR0006685|BWBR0001833|BWBR0001840|BWBR0002221| BWBR0031788|BWBR0003968|BWBR0005034|BWBR0002154| BWBR0012287|BWBV0001506|BWBR0006358|BWBR0003270| BWBR0003738|BWBR0005416|BWBR0005848|BWBR0022704| BWBR0011470|BWBR0006330|BWBR0002170|BWBR0001941| BWBR0002367|BWBR0002368|BWBR0002757|BWBR0002629| BWBR0002375|BWBR0005645|BWBR0009709|BWBR0020449| BWBR0005674|BWBV0001017|BWBR0018450|BWBV0003149| BWBR0013605|BWBR0002126|BWBR0002467|BWBV0005225| BWBR0009225|BWBR0013409|BWBR0006299|BWBR0006763| BWBR0003391|BWBR0005366|BWBR0009398|BWBR0006251| BWBV0005096|BWBR0001827|BWBR0012810|BWBR0002320| BWBR0006622|BWBR0016664|BWBV0001021|BWBR0011222| BWBR0011453|BWBR0002524|BWBR0004581|BWBR0009726| BWBR0010346|BWBR0008498|BWBR0020611|BWBR0021505| BWBR0006368|BWBR0004619|BWBR0007671|BWBR0005212| BWBR0011695|BWBR0002471|BWBR0005252|BWBR0002548| BWBR0004471|BWBR0001886|BWBR0004452|BWBR0011788| BWBR0006297|BWBV0002030|BWBR0009408|BWBR0001888| BWBR0006705|BWBR0002505|BWBR0011353|BWBV0004027| BWBR0001950|BWBR0002032|BWBR0001830|BWBV0003264| BWBR0008804|BWBR0004028|BWBR0017329|BWBV0003441| BWBR0014809|BWBR0013310|BWBR0014397|BWBV0001507| BWBR0006517|BWBR0001947|BWBR0002448|BWBR0001869| BWBV0006026|BWBR0002267|BWBR0013513|BWBR0006001| BWBR0008657|BWBR0003482|BWBR0010424|BWBR0003384| BWBR0002460|BWBR0004045|BWBR0001937|BWBR0007791| BWBR0013918|BWBR0013604|BWBV0001800|BWBR0020674| BWBR0029361|BWBR0012125|BWBV0003487|BWBR0007523)")) 8 9 }UNION{ 10 11 ?ref mo:cites ?bwbs. 12 FILTER (regex(str(?ref ),"(?:BWBR0011825|BWBR0009876|
BWBR0005537|BWBV0001000|BWBR0015703|BWBR0011823| BWBR0007333|BWBV0002508|BWBR0017646|BWBV0004701| BWBR0002638|BWBR0001903|BWBR0005181|BWBV0001002| BWBR0006723|BWBR0001854|BWBR0012002|BWBR0017959| BWBR0005310|BWBR0005290|BWBR0007149|BWBR0011468| BWBR0006685|BWBR0001833|BWBR0001840|BWBR0002221| BWBR0031788|BWBR0003968|BWBR0005034|BWBR0002154| BWBR0012287|BWBV0001506|BWBR0006358|BWBR0003270| BWBR0003738|BWBR0005416|BWBR0005848|BWBR0022704| BWBR0011470|BWBR0006330|BWBR0002170|BWBR0001941| BWBR0002367|BWBR0002368|BWBR0002757|BWBR0002629| BWBR0002375|BWBR0005645|BWBR0009709|BWBR0020449| BWBR0005674|BWBV0001017|BWBR0018450|BWBV0003149| BWBR0013605|BWBR0002126|BWBR0002467|BWBV0005225| BWBR0009225|BWBR0013409|BWBR0006299|BWBR0006763| BWBR0003391|BWBR0005366|BWBR0009398|BWBR0006251| BWBV0005096|BWBR0001827|BWBR0012810|BWBR0002320| BWBR0006622|BWBR0016664|BWBV0001021|BWBR0011222| BWBR0011453|BWBR0002524|BWBR0004581|BWBR0009726| BWBR0010346|BWBR0008498|BWBR0020611|BWBR0021505| BWBR0006368|BWBR0004619|BWBR0007671|BWBR0005212| BWBR0011695|BWBR0002471|BWBR0005252|BWBR0002548| BWBR0004471|BWBR0001886|BWBR0004452|BWBR0011788| BWBR0006297|BWBV0002030|BWBR0009408|BWBR0001888| BWBR0006705|BWBR0002505|BWBR0011353|BWBV0004027| BWBR0001950|BWBR0002032|BWBR0001830|BWBV0003264| BWBR0008804|BWBR0004028|BWBR0017329|BWBV0003441| BWBR0014809|BWBR0013310|BWBR0014397|BWBV0001507| BWBR0006517|BWBR0001947|BWBR0002448|BWBR0001869| BWBV0006026|BWBR0002267|BWBR0013513|BWBR0006001| BWBR0008657|BWBR0003482|BWBR0010424|BWBR0003384| BWBR0002460|BWBR0004045|BWBR0001937|BWBR0007791| BWBR0013918|BWBR0013604|BWBV0001800|BWBR0020674| BWBR0029361|BWBR0012125|BWBV0003487|BWBR0007523)")) 13 } 14 }
B
Example of a ECLI file after parsing
1 <?xml version=’1.0’ encoding=’utf8’?> 2 <open-rechtspraak xmlns:ns1="http://purl.org/dc/terms/" xmlns:ns2 ="http://www.w3.org/2000/01/rdf-schema#" xmlns:ns3="http:// psi.rechtspraak.nl/" xmlns:ns4="bwb-dl" xmlns:ns5="http://www .rechtspraak.nl/schema/rechtspraak-1.0" xmlns:rdf="http://www .w3.org/1999/02/22-rdf-syntax-ns#"> 3 <rdf:RDF> 4 <rdf:Description> 5 <ns1:identifier>ECLI:NL:RVS:2012:BY7272</ns1: identifier> 6 <ns1:format>text/xml</ns1:format> 7 <ns1:accessRights>public</ns1:accessRights> 8 <ns1:modified>2014-04-13T14:14:46</ns1:modified> 9 <ns1:issued ns2:label="Publicatiedatum">2013-04-05</ ns1:issued> 10 <ns1:publisher resourceIdentifier="http://rechtspraak. nl/">Raad voor de Rechtspraak</ns1:publisher> 11 <ns1:language>nl</ns1:language> 12 <ns1:replaces ns2:label="Vervangt">BY7272</ns1: replaces> 13 <ns1:creator resourceIdentifier="http://standaarden. overheid.nl/owms/terms/Raad_van_State" scheme=" overheid.RechterlijkeMacht" ns2:label="Instantie"> Raad van State</ns1:creator>14 <ns1:date ns2:label="Uitspraakdatum">2012-12-19</ns1: date>
15 <ns3:zaaknummer ns2:label="Zaaknr">201100481/1/V4</ns3 :zaaknummer>
16 <ns1:type resourceIdentifier="http://psi.rechtspraak. nl/uitspraak" rdf:language="nl">Uitspraak</ns1:type
> 17 <ns3:procedure resourceIdentifier="http://psi. rechtspraak.nl/procedure#hogerBeroep" rdf:language= "nl" ns2:label="Procedure">Hoger beroep</ns3: procedure> 18 <ns1:coverage>NL</ns1:coverage> 19 <ns1:subject resourceIdentifier="http://psi. rechtspraak.nl/rechtsgebied# bestuursrecht_vreemdelingenrecht" ns2:label=" Rechtsgebied">Bestuursrecht; Vreemdelingenrecht</ ns1:subject> 20 <ns1:references ns4:resourceIdentifier="1.0:v:BWB: BWBR0005537" ns2:label="Wetsverwijzing">Algemene wet bestuursrecht</ns1:references> 21 <ns1:references ns4:resourceIdentifier="1.0:v:BWB: BWBR0005537&artikel=4:6" ns2:label="
:references> 22 <ns1:references ns4:resourceIdentifier="1.0:v:BWB: BWBR0011823" ns2:label="Wetsverwijzing"> Vreemdelingenwet 2000</ns1:references> 23 <ns1:references ns4:resourceIdentifier="1.0:v:BWB: BWBR0011823&artikel=83" ns2:label=" Wetsverwijzing">Vreemdelingenwet 2000 83</ns1: references> 24 <ns1:hasVersion resourceIdentifier="http://psi. rechtspraak.nl/vindplaats" ns2:label="Vindplaatsen" > 25 <rdf:list> 26 <rdf:li>Rechtspraak.nl</rdf:li> 27 <rdf:li>JV 2013/72</rdf:li> 28 </rdf:list> 29 </ns1:hasVersion> 30 <ns1:references metaLexResourceIdentifier="http://doc. metalex.eu/id/BWBR0005537/artikel/8%3A54" ns4: resourceIdentifier="1.0:v:BWB:BWBR0005537& artikel=8%3A54" ns2:label="Wetsverwijzing"> 31 <ns1:string> artikel 8:54, eerste lid, van de
Algemene wet bestuursrecht (hierna: de Awb)</ ns1:string> 32 </ns1:references> 33 <ns1:references metaLexResourceIdentifier="http://doc. metalex.eu/id/BWBR0011823/artikel/83" ns4: resourceIdentifier="1.0:v:BWB:BWBR0011823& artikel=83" ns2:label="Wetsverwijzing">
34 <ns1:string> artikel 83 van de Vreemdelingenwet 2000 (hierna: de Vw 2000)</ns1:string> 35 </ns1:references> 36 <ns1:references metaLexResourceIdentifier="http://doc. metalex.eu/id/BWBR0011823/artikel/83" ns4: resourceIdentifier="1.0:v:BWB:BWBR0011823& artikel=83" ns2:label="Wetsverwijzing">
37 <ns1:string> artikel 83, eerste lid, van de Vw 2000 </ns1:string> 38 </ns1:references> 39 <ns1:references metaLexResourceIdentifier="http://doc. metalex.eu/id/BWBR0011823/artikel/64" ns4: resourceIdentifier="1.0:v:BWB:BWBR0011823& artikel=64" ns2:label="Wetsverwijzing">
40 <ns1:string> artikel 64 van de Vw 2000 </ns1:string > 41 </ns1:references> 42 </rdf:Description> 43 <rdf:Description rdf:about="http://deeplink.rechtspraak.nl /uitspraak?id=ECLI:NL:RVS:2012:BY7272"> 44 <ns1:identifier>http://deeplink.rechtspraak.nl/ uitspraak?id=ECLI:NL:RVS:2012:BY7272</ns1:
identifier> 45 <ns1:format>text/html</ns1:format> 46 <ns1:accessRights>public</ns1:accessRights> 47 <ns1:modified>2013-04-05T11:20:10</ns1:modified> 48 <ns1:issued ns2:label="Publicatiedatum">2012-12-27</ ns1:issued> 49 <ns1:publisher resourceIdentifier="http://rechtspraak. nl/">Raad voor de Rechtspraak</ns1:publisher> 50 <ns1:language>nl</ns1:language>
51 <ns1:title rdf:language="nl">ECLI:NL:RVS:2012:BY7272 Raad van State , 19-12-2012 / 201100481/1/V4</ns1: title>
52 <ns1:abstract resourceIdentifier="../../rs: inhoudsindicatie" />
53 </rdf:Description> 54 </rdf:RDF>
55 <ns5:inhoudsindicatie id="ECLI:NL:RVS:2012:BY7272:INH" lang=" nl" xml:space="preserve">
56 <ns5:para>DUBLINVERORDENING. Artikel 3 lid 2. Polen. Psychische gesteldheid had bij beoordeling beroep moeten worden betrokken met toepassing van artikel 83 van de Vw 2000. BMA ten onrechte niet ingeschakeld.</ ns5:para>
57 </ns5:inhoudsindicatie>
58 <ns5:uitspraak id="ECLI:NL:RVS:2012:BY7272:DOC" lang="nl" xml: space="preserve">
59 <ns5:parablock>
60 <ns5:para>201100481/1/V4.</ns5:para>
61 <ns5:para>Datum uitspraak: 19 december 2012</ns5:para> 62 </ns5:parablock>
63 <ns5:para /> 64 <ns5:parablock>
65 <ns5:para>Raad van State</ns5:para> 66 <ns5:para>AFDELING</ns5:para>
67 <ns5:para>BESTUURSRECHTSPRAAK</ns5:para> 68 </ns5:parablock>
69 <ns5:para /> 70 <ns5:parablock>
71 <ns5:para>Uitspraak met toepassing van artikel 8:54, eerste lid, van de Algemene wet bestuursrecht ( hierna: de Awb) op het hoger beroep van:</ns5:para> 72 <ns5:para>[naam],</ns5:para>
73 <ns5:para>appellant,</ns5:para> 74 <ns5:para>tegen de uitspraak van de
voorzieningenrechter van de rechtbank ’s Gravenhage , nevenzittingsplaats Maastricht, van 15 december 2010 in zaak nrs. 10/37556 en 10/37557 in het geding tussen:</ns5:para>
75 <ns5:para>de vreemdeling</ns5:para> 76 <ns5:para>en</ns5:para>
77 <ns5:para>de minister voor Immigratie en Asiel.</ns5: para> 78 </ns5:parablock> 79 <ns5:para/> 80 <ns5:parablock> 81 <ns5:para>Procesverloop</ns5:para>
82 <ns5:para>Bij besluit van 28 oktober 2010 heeft de minister een aanvraag van de vreemdeling om hem een
verblijfsvergunning asiel voor bepaalde tijd te verlenen afgewezen. Dit besluit is aangehecht.</ns5 :para>
83 <ns5:para>Bij uitspraak van 15 december 2010 heeft de voorzieningenrechter, voor zover thans van belang, het daartegen door de vreemdeling ingestelde beroep
ongegrond verklaard. Deze uitspraak is aangehecht. </ns5:para>
84 <ns5:para>Tegen deze uitspraak heeft de vreemdeling hoger beroep ingesteld. Het hogerberoepschrift is aangehecht. </ns5:para>
85 <ns5:para>De minister, thans: de staatssecretaris van Veiligheid en Justitie, heeft een verweerschrift ingediend.</ns5:para>
86 <ns5:para>Vervolgens is het onderzoek gesloten.</ns5: para>
87 </ns5:parablock> 88 <ns5:para /> 89 <ns5:parablock>
90 <ns5:para>Overwegingen</ns5:para>
91 <ns5:para>1. Onder de staatssecretaris wordt tevens verstaan: diens rechtsvoorgangers.</ns5:para> 92 <ns5:para>2. Uit de jurisprudentie van de Afdeling (
uitspraak van 6 maart 2008 in zaak nr. 200706839/1, www.raadvanstate.nl) vloeit voort dat, indien een bestuursorgaan na een eerder afwijzend besluit een besluit van gelijke strekking neemt, door het instellen van beroep tegen het laatste besluit niet
kan worden bereikt dat de bestuursrechter dat besluit toetst, als ware het een eerste afwijzing. Dit uitgangspunt geldt niet alleen voor besluiten genomen naar aanleiding van een nieuwe aanvraag, maar ook voor besluiten op een verzoek om terug te komen van een al dan niet op aanvraag genomen besluit (uitspraak van de Afdeling van 4 mei 2005 in zaak nr. 200406320/1, www.raadvanstate.nl). Slechts indien en voor zover in de bestuurlijke fase nieuw gebleken feiten of veranderde
omstandigheden zijn aangevoerd, dan wel uit het aldus aangevoerde kan worden afgeleid dat zich een relevante wijziging van het recht heeft voorgedaan,
ervan en de wijze waarop het tot stand is gekomen toetsen.</ns5:para>
93 <ns5:para>2.1. Onder nieuw gebleken feiten of
veranderde omstandigheden moeten worden begrepen feiten of omstandigheden die na het eerdere besluit
zijn voorgevallen of die niet v´o´or dat besluit konden en derhalve behoorden te worden aangevoerd, alsmede bewijsstukken van reeds eerder gestelde feiten of omstandigheden, die niet v´o´or het nemen van het eerdere besluit konden en derhalve
behoorden te worden overgelegd. Is hieraan voldaan, dan doen zich niettemin geen feiten of
omstandigheden voor die een hernieuwde toetsing rechtvaardigen, indien op voorhand is uitgesloten dat hetgeen alsnog is aangevoerd of overgelegd aan het eerdere besluit kan afdoen.</ns5:para>
94 <ns5:para>2.2. De moeder van de vreemdeling heeft eerder, op 27 november 2009, mede voor de vreemdeling een aanvraag tot verlening van een verblijfsvergunning asiel voor bepaalde tijd ingediend. Bij besluit van 17 mei 2010 heeft de staatssecretaris deze aanvraag afgewezen, omdat ingevolge Verordening (EG) 343/2003 van de Raad van
18 februari 2003 tot vaststelling van de criteria en instrumenten om te bepalen welke lidstaat verantwoordelijk is voor de behandeling van een asielverzoek dat door een onderdaan van een derde land bij een van de lidstaten wordt ingediend (PB 2003 L 050; hierna: de Verordening) Polen
verantwoordelijk is voor de behandeling van het asielverzoek. </ns5:para>
95 <ns5:para> Op 20 oktober 2010 heeft de vreemdeling een aanvraag tot verlening van een verblijfsvergunning asiel voor bepaalde tijd ingediend. De
staatssecretaris heeft deze aanvraag bij besluit van 28 oktober 2010 afgewezen, omdat ingevolge de Verordening Polen verantwoordelijk is voor de behandeling van het asielverzoek.</ns5:para> 96 <ns5:para> Het besluit van 28 oktober 2010 is van
gelijke strekking als dat van 17 mei 2010, voor zover de staatssecretaris bij het laatstvermelde besluit de voor de vreemdeling ingediende aanvraag heeft afgewezen. Derhalve is op het tegen het nieuwe besluit gerichte beroep het hiervoor onder 2. weergegeven beoordelingskader van toepassing ( vergelijk de uitspraak van de Afdeling van 23 april
2008 in zaak nr. 200706841/1; www.raadvanstate.nl) .</ns5:para>
97 <ns5:para>2.3. De vreemdeling klaagt in grief 1 dat de voorzieningenrechter ten onrechte heeft overwogen
dat de psychische gezondheidstoestand van de vreemdeling geen nieuw gebleken feit of veranderde omstandigheid is, omdat de vreemdeling die reeds in
de bestuurlijke besluitvormingsprocedure naar voren had moeten brengen. Door aldus te overwegen heeft de voorzieningenrechter volgens de
vreemdeling miskend dat hij, nu zijn psychische gesteldheid eerst sinds 28 oktober 2010 is gewijzigd, de stukken ter staving van zijn gewijzigde psychische gesteldheid niet in de bestuurlijke fase voorafgaande aan het besluit van 28 oktober 2010 kon overleggen. De
voorzieningenrechter heeft deze stukken dan ook ten onrechte niet krachtens artikel 83 van de
Vreemdelingenwet 2000 (hierna: de Vw 2000) bij de beoordeling van het beroep betrokken, aldus de vreemdeling.</ns5:para>
98 <ns5:para>2.4. De vreemdeling heeft in beroep betoogd dat de staatssecretaris, krachtens artikel 3, tweede lid, van de Verordening gebruik had moeten maken van zijn bevoegdheid om de asielaanvraag aan zich te trekken, aangezien als gevolg van de gewijzigde psychische gesteldheid van de
vreemdeling de feitelijke overdracht aan Polen zal leiden tot een acute medische noodsituatie. Ter staving van dit betoog heeft de vreemdeling een e-mailbericht van 29 oktober 2010 van een medewerker van de stichting Nidos (hierna: Nidos), een e-mailbericht van 5 november 2010 van een medewerker van het Centraal Orgaan opvang asielzoekers (hierna : het COa) en een brief van 18 november 2010 van zijn behandelend psycholoog (hierna: de psycholoog)
overgelegd.</ns5:para>
99 <ns5:para>2.5. Het e-mailbericht van 29 oktober 2010, gericht aan de gemachtigde van de vreemdeling, bevat een passage uit een verslag van </ns5:para> 100 <ns5:para>28 oktober 2010 dat een medewerker van het
COa heeft opgesteld en dat aan een medewerker van Nidos is gericht. Deze passage vermeldt dat de vreemdeling op 28 oktober 2010 door medewerkers van
het COa op de grond in zijn kamer is aangetroffen, met verwondingen aan zijn linker onderarm. Aan ´e´en van de medewerkers van het COa heeft de
vreemdeling te kennen gegeven zelfmoord te willen plegen. </ns5:para>
101 <ns5:para>In het e-mailbericht van 5 november 2010 deelt een medewerker van het COa de gemachtigde van
de vreemdeling mede dat de vreemdeling op 4 november 2010 een gesprek heeft gehad met de psycholoog.</ns5:para>
102 <ns5:para>In de brief van 18 november 2010 vermeldt de psycholoog dat hij op basis van een psychologisch onderzoek een zeer acute mate van su¨ıcidaliteit bij
de vreemdeling heeft vastgesteld. Voorts vermeldt de psycholoog dat hij het risico op su¨ıcide
momenteel zeer hoog inschat, omdat bij de vreemdeling een zeer rationele afweging heeft plaatsgevonden, hij emotioneel afscheid heeft genomen van door hem in vertrouwen genomen personen
in zijn directe omgeving en hij een nauwkeurig uitgewerkt plan heeft voor hoe hij su¨ıcide wil plegen. De psycholoog licht voorts toe dat de vreemdeling voldoet aan alle criteria om een complexe posttraumatische stressstoornis volgens het internationale classificatie instrument Diagnostic and Statistical Manual of Mental Disorders, versie IV-r (lees: versie IV-TR), vast te stellen. Ten slotte concludeert de psycholoog dat de vreemdeling acuut verdere psychiatrische hulp nodig heeft ter preventie van su¨ıcide en een mogelijk gedwongen terugkeer naar Polen erg riskant
is, omdat het met grote waarschijnlijkheid op zeer korte termijn zal leiden tot effectieve su¨ıcide.</ ns5:para>
103 <ns5:para>2.6. Uit de door de vreemdeling overgelegde stukken blijkt dat de door de psycholoog
vastgestelde psychische klachten van de vreemdeling , waaronder de zeer acute mate van su¨ıcidaliteit, zich op 28 oktober 2010 hebben geopenbaard. Gelet hierop en nu het besluit van gelijke strekking op 28 oktober 2010 is uitgereikt, heeft de
voorzieningenrechter ten onrechte overwogen dat de vreemdeling zijn psychische gesteldheid reeds in de
bestuurlijke fase had moeten aanvoeren. Nu bovendien de psychische klachten, zoals door de psycholoog vastgesteld, na het nemen van het eerdere besluit zijn voorgevallen en niet op voorhand is uitgesloten dat die klachten kunnen afdoen aan het eerdere besluit, is de psychische gesteldheid van de vreemdeling, zoals door de psycholoog in de brief van 18 november 2010 vastgesteld, een nieuw gebleken feit dan wel een veranderde omstandigheid, die krachtens artikel 83,
eerste lid, van de </ns5:para>
104 <ns5:para>Vw 2000 bij de beoordeling van het beroep moet worden betrokken. De grief slaagt.</ns5:para> 105 <ns5:para>3. Het hoger beroep is kennelijk gegrond.
Hetgeen in grief 2 is aangevoerd, behoeft geen bespreking. De aangevallen uitspraak moet worden vernietigd. Doende hetgeen de rechtbank zou behoren
te doen, wordt als volgt overwogen.</ns5:para> 106 <ns5:para>4. De staatssecretaris heeft zich in het in
het besluit van 28 oktober 2010 ingelaste voornemen op het standpunt gesteld dat thans geen feiten of omstandigheden aanwezig zijn op grond waarvan Nederland krachtens artikel 3, tweede lid, van de Verordening het asielverzoek in behandeling zou moeten nemen.</ns5:para>
107 <ns5:para> Over de psychische gesteldheid van de vreemdeling heeft de staatssecretaris zich ter zitting bij de voorzieningenrechter op het
standpunt gesteld dat deze niet leidt tot een ander standpunt dan in het besluit van 28 oktober 2010 verwoord. Daartoe heeft de staatssecretaris aangevoerd dat medische voorzieningen in Polen in beginsel aanwezig zijn, de vreemdeling niet aannemelijk heeft gemaakt dat in zijn geval aanwijzingen bestaan dat dit uitgangspunt niet opgaat en dat het de vreemdeling vrijstaat een beroep op artikel 64 van de Vw 2000 te doen. </ns5: para>
108 <ns5:para>4.1. Volgens paragraaf C3/2.3.6.4. van de Vreemdelingencirculaire 2000 (hierna: de Vc 2000) kan de staatssecretaris, indien een vreemdeling op basis van bijzondere, individuele omstandigheden aannemelijk maakt dat het overdragen van de vreemdeling aan de verantwoordelijke lidstaat van een onevenredige hardheid getuigt, ook in
dergelijke individuele gevallen gebruik maken van de bevoegdheid van artikel 3, tweede lid, van de Verordening. Welke aspecten in dit kader een rol kunnen spelen, is niet zonder meer te duiden nu het
met name zal gaan om de omstandigheden van het geval en in hoeverre deze bijzondere omstandigheden
kunnen leiden tot de conclusie dat voortzetting van de Dublinprocedure onevenredig hard is. Daarbij
is de enkele aanwezigheid van medische aspecten niet voldoende om te spreken van bijzondere
omstandigheden, omdat de medische voorzieningen in beginsel vergelijkbaar worden verondersteld tussen de lidstaten en het interstatelijk
vertrouwensbeginsel er eveneens van uitgaat dat de voorzieningen in de lidstaten - indien ge¨ındiceerd - ook ter beschikking staan voor de Dublinclaimant.
Dit lijdt slechts uitzondering indien een
vreemdeling aannemelijk maakt dat dit uitgangspunt in zijn of haar geval niet opgaat, aldus de Vc 2000. </ns5:para>
109 <ns5:para>4.2. Volgens voormeld beleid kan de
aanleiding geven voor toepassing van artikel 3, tweede lid, van de Verordening, indien de
vreemdeling aannemelijk heeft gemaakt dat het in het beleid weergegeven uitgangspunt in zijn geval niet opgaat.</ns5:para>
110 <ns5:para>4.3. In de brief van 18 november 2010
concludeert de psycholoog dat de vreemdeling acuut verdere psychiatrische hulp nodig heeft ter
preventie van su¨ıcide en een mogelijk gedwongen terugkeer naar Polen erg riskant is, omdat deze met
grote waarschijnlijkheid op zeer korte termijn zal leiden tot effectieve su¨ıcide. Gelet op deze informatie heeft de staatssecretaris, alvorens een besluit op de aanvraag te nemen, in dit geval het Bureau Medische Advisering ten onrechte niet laten onderzoeken of de psychische gesteldheid van de vreemdeling aan de overdracht aan Polen in de weg staat en, zo nee, met inachtneming van welke
vereisten deze overdracht moet plaatsvinden. Nu dit niet is gebeurd, heeft de staatssecretaris zijn standpunt dat de psychische gesteldheid van de vreemdeling geen grond biedt voor het oordeel dat hij de behandeling van het asielverzoek van de vreemdeling krachtens artikel 3, tweede lid, van de
Verordening aan zich moet trekken, ondeugdelijk gemotiveerd. </ns5:para>
111 <ns5:para>5. Het beroep tegen het besluit van 28 oktober 2010 is gegrond. Het besluit komt voor vernietiging in aanmerking wegens strijd met artikel 3:2 en 3:46 van de Awb. </ns5:para> 112 <ns5:para>6. De staatssecretaris moet op na te melden
wijze tot vergoeding van de proceskosten worden veroordeeld.</ns5:para>
113 </ns5:parablock> 114 <ns5:para/> 115 <ns5:parablock>
116 <ns5:para>Beslissing</ns5:para>
117 <ns5:para>De Afdeling bestuursrechtspraak van de Raad van State</ns5:para>
118 <ns5:para>Recht doende in naam der Koningin:</ns5:para >
119 <ns5:para>I. verklaart het hoger beroep gegrond;</ns5: para>
120 <ns5:para>II. vernietigt de uitspraak van de
voorzieningenrechter van de rechtbank ’s-Gravenhage , nevenzittingsplaats Maastricht, van 15 december 2010 in zaak nr. 10/37556;</ns5:para>
121 <ns5:para>III. verklaart het in die zaak ingestelde beroep gegrond;</ns5:para>
voor Immigratie en Asiel van 28 oktober 2010, kenmerk 0911-27-1495;</ns5:para>
123 <ns5:para>V. veroordeelt de staatssecretaris van Veiligheid en Justitie tot vergoeding van bij de vreemdeling in verband met de behandeling van beroep en hoger beroep opgekomen proceskosten tot een bedrag van e1.311,00 (zegge: dertienhonderdelf
euro), geheel toe te rekenen aan door een derde beroepsmatig verleende rechtsbijstand.</ns5:para> 124 </ns5:parablock>
125 <ns5:para />
126 <ns5:para>Aldus vastgesteld door mr. H.G. Lubberdink, voorzitter, en mr. R. van der Spoel en mr. N. Verheij,
leden, in tegenwoordigheid van mr. N. Walcott Oliai, ambtenaar van staat.</ns5:para>
127 <ns5:para /> 128 <ns5:parablock>
129 <ns5:para>w.g. Lubberdink</ns5:para>
130 <ns5:para>voorzitter w.g. Walcott-Oliai</ns5:para> 131 <ns5:para>ambtenaar van staat</ns5:para>
132 </ns5:parablock> 133 <ns5:para />
134 <ns5:para>Uitgesproken in het openbaar op 19 december 2012 </ns5:para>
135 <ns5:para /> 136 <ns5:parablock>
137 <ns5:para>555-633.</ns5:para>
138 <ns5:para>Verzonden: 19 december 2012</ns5:para> 139 </ns5:parablock>
140 <ns5:para /> 141 <ns5:parablock>
142 <ns5:para>Voor eensluidend afschrift,</ns5:para> 143 <ns5:para>de secretaris van de Raad van State,</ns5:
para> 144 </ns5:parablock> 145 <ns5:para/> 146 <ns5:para>mr. H.H.C. Visser</ns5:para> 147 </ns5:uitspraak> 148 </open-rechtspraak>
